The purpose of this paper is to study the relevance of an architecture where DiffServ, MPLS and network management would cooperate to overcome the challenges for QoS-capable IP networks.

Aeronautical Telecommunication Network (ATN) has been developed by the International Civil Aviation Organization to integrate Air-Ground and Ground-Ground data communication for aeronautical applications into a single network serving Air Traffic Control and Aeronautical Operational Communications. To carry time critical information required for aeronautical applications, ATN provides different Quality of Services (QoS) to applications. ATN has been designed as a standalone network with its own protocols which requires building an expensive separate network for ATN. However, the cost of building ATN can be reduced if it can run over a public network such as the Internet. Although the current Internet does not provide QoS, Internet Engineering Task Force (IETF) is standardizing the Differentiated Services (DiffServ) network to provide differential QoS to users of next generation data networks. The objective of this paper is to investigate the possibility of providing QoS to ATN applications when it runs over the DiffServ network in the next generation Internet. Our results show that the QoS requirements of ATN applications can be successfully provided when they run over a DiffServ backbone in the next generation Internet.

In this paper, we discuss transport layer protocols to support real-time data transmission overnetworks using Differentiated Services (Diffserv) where bandwidth is guaranteed. Assuming that an application transmits streaming video data over a Diffserv network, we have performed simulation experiments to evaluate performance of real-time transmission using standard internet protocols, UDP and TCP as a transport layer protocol. The application is designed to buffer a portion of the data on the receiver before starting the playback and to operate in a streaming mode. Consequently, when packets are lost over Diffserv networks, TCP shows better performance than UDP. Simulation results show that employment of SACK option for TCP can recover lost packets efficiently and improve performance of real-time transmission. Currently, a growing number of real-time media applications over Internet are being implemented using UDP, because reliable delivery mechanism of TCP can obstruct real-time data transmission that is timing critical. However, in our experimental environment, retransmission and congestion control mechanism of TCP make use of bandwidth guaranteed by Diffserv, and help to provide more reliable delivery of real-time data than UDP without serious impacts on time-sensitive delivery. Therefore, despite of the general belief that TCP is not suitable for time-sensitive delivery, in our experiment, we show that TCP is a transport layer protocol suitable for real-time data transmission over Diffserv networks.

Quality-of-service (QoS) requirements for the timely delivery of real-time multimedia raise new challenges for today's networks. A key component of QoS is QoS routing which allows the selection of network routes with sufficient resources for requested QoS parameters. Most proposed techniques to compute QoS routes require dynamic update of link-state information. Given the growing size of internets, it is becoming increasingly difficult to gather up-to-date state information. We propose a new technique to compute QoS routes in a fast and efficient manner without any need for dynamic updates. Our method, known as Selective Flooding, checks the state of the links on a set of pre-computed routes from the source to the destination in parallel and based on this information computes the best route and then reserves resources. We implemented Selective Flooding on a QoS routing simulator and evaluated the performance of Selective Flooding compared to source routing for a variety of network parameters. We find Selective Flooding consistently outperforms source routing in terms of call-blocking rate and outperforms source routing in terms of network overhead for some network conditions. The contributions of this work include the design of a new QoS routing algorithm, Selective Flooding, extensive evaluation of Selective Flooding under a variety of network conditions and a working simulator for future research.

New services are introduced in the broadband network such video or multimedia. These services have some quality of service constraint such as transit delay, cell loss probability. To improve the quality of service, it is important to optimize the bandwidth utilization. In this study we focus on this problem, and our major contribution is to provide quality of service by the development of a new rerouting algorithm. All existing routing algorithms optimize at the time of call establishment. The originality of our contribution is to optimize routing before and after call establishment. To show the benefit of our algorithm we do a performance evaluation using a new technique called SAN: Stochastic Automata Networks, which is adapted to our problem.

In this paper we present an inter-vehicle communication protocol based on geographical routing. Our proposed protocol enables point-to-point communications between cars in highway. The protocol is designed for highway travelers but can be used in any mobile ad-hoc network. The highway is divided in virtual cells, which moves as the vehicles moves. The cell members choose a center that will behave for a certain time interval as a Base Station. Every node has its geographical position given by Global Positioning System (GPS). When a source node has a message for a destination node, forwards it to its Cell Center. Then the message is forwarded through the other Cell Centers. The Cell Centers first verifies if the destination node belongs to their cell. Finally the destination Cell Center will send the message to the destination node. Our simulation results show that our proposed protocol improves the network utilization compared to existing inter-vehicles protocols. The protocol can be used to implement differentiated mobile services and message prioritization.

The thrust of this study is to construct an MPLS test-bed using open hardware and software and later use this test-bed for experimenting various traffic engineering options available with MPLS. We have constructed a test-bed using Pentium PCs and Linux and used this test-bed to try a well-known MPLS traffic engineering feature of separating flows into multiple trunks. The purpose of this separation is to experimentally assess the quality of service benefits we can expect from MPLS networks. We experiment with a hierarchical separation structure where the first level of separation is based on the transport layer used (TCP or UDP). This high level separation prevents UDP from unfairly grabbing the available bandwidth from TCP applications. Since different applications using TCP have different needs, e.g., some might want high bandwidth while others are more sensitive to packet delay, the second level of separation is based on application type. The results from our open experimental test-bed confirms the intuitive results by demonstrating better QoS with the flow separation in the MPLS core.

The Label Switching technology was developed in order to provide a high speed and large packet processing capabilities in Internet networks using a cut-through technique. Research effort has concentrated on the issue of number of labels used and its minimization. There are mainly two label mapping policies. One is the data-driven label mapping policy and the other is the control-driven label mapping policy. In this paper, a comparison between both conventional label mapping policies is shown. It is also evaluated the required number of labels under each of these two label mapping policies and the cut-through ratio of each one. The evaluation shows that the data-driven and the control-driven label mapping policies require a large number of labels. In order to reduce the number of labels required and increase the cut-through ratio, it is proposed a label mapping policy that is a combination of the two label mappings above. This is a control-driven label mapping that has the soft-state characteristic of the data-driven one. The evaluation shows that the proposed label mapping policy requires less labels than the others conventional label mapping policies and increases the cut-through ratio to the maximum allowed while offering the advantages of being a control-driven label mapping.

In this work, we present online algorithms for dynamic routing bandwidth guaranteed label switched paths (LSPs) where LSP set-up requests (in terms of a pair of ingress and egress routers as well as its bandwidth requirement) arrive one by one and there is no a priori knowledge regarding future LSP set-up requests. In addition, we consider rerouting of LSPs in this work. Rerouting of LSPs has not been well studied in previous work on LSP routing. The need of LSP rerouting arises in a number of ways: occurrence of faults (link and/or node failures), re-optimization of existing LSPs' routes to accommodate traffic fluctuation, requests with higher priorities, and so on. We formulate the bandwidth guaranteed LSP routing with rerouting capability as a multi-commodity flow problem. The solution to this problem is used as the benchmark for comparing other computationally less costly algorithms studied in this paper. Furthermore, to more efficiently utilize the network resources, we propose online routing algorithms which route bandwidth demands over multiple paths at the ingress router to satisfy the customer requests while providing better service survivability. Traffic splitting and distribution over the multiple paths are carefully handled using table-based hashing schemes while the order of packets within a flow is preserved. Preliminary simulations are conducted to show the performance of different design choices and the effectiveness of the rerouting and multi-path routing algorithms in terms of LSP set-up request rejection probability and bandwidth blocking probability.

Providing Quality of Service (QoS) capabilities in IP networks is the focus of much research in the Internet community today. The Intserv architecture proposed by IETF for supporting QoS in IP networks is known not to scale well to large networks due to the per-flow mechanisms it uses. While connection-less approach has been the central design paradigm of IP networks, many researches today believe connection-oriented mechanisms similar to virtual-circuits in ATM must be incorporated in IP networks in order to provide scalable QoS. Towards this end, the MPLS (Multi-Protocol Label Switching) technology has been put forward by the IETF. A crucial scaling issue with MPLS is that large number of LSPs (Label Switched Paths) may be required in the routers resulting in increase of state size in the routers. In this paper, we introduce the notion of Label-Switched Multipaths (LSMPs) and propose a simple technique for aggregating LSP into LSMPs such that the number of labels required in the routers is significantly reduced. Based on LSMPs we describe an architecture for providing deterministic guarantees that is far more scalable than architectures based on simple LSPs or those that use only multipoint-to-point LSP aggregation.

Advances in optical networking technologies are enabling new topologies to be considered that differ considerably from those prevalent in existing classical communication networks. Some of these new topologies include next-generation abstractions of SONET/SDH multi-ring networks and more recently, mesh topology-based IP-over-WDM applications for long-haul and regional/metro optical networks. A key requirement in emerging optical networks is that of light path routing, which poses many topology-specific constraints. In this paper we consider embedding of the light paths in full-mesh network into the existing fiber-optic cables topologies (host topology) in such a way that the resulting embedding will have minimum fiber-bundle (fiber link width) in the host topology over all other embeddings. We present solutions of this problem when the host fiber-optic cables topologies are ring (cycle), tree, union of rings, wheels. For an arbitrary host network topology the problem appears to be a more difficult graph-embedding problem. The paper also deals with restoration and protections in the optical mesh WDM networks, which are based on 2-factorizations.

The Internet Protocol (IP) has served global networks well, providing a standardized method to transmit data among many disparate systems. But IP is designed for simplicity, and only enables a `best effort' service that can be subject to delays and loss of data. For data networks, this is an acceptable trade-off. In the emerging world of convergence, driven by new applications such as video streaming and IP telephony, minimizing latency and packet loss as well as jitter can be critical. Simply increasing the size of the IP network `pipe' to meet those demands is not always sufficient. In this environment, vendors and standards bodies are endeavoring to create technologies and techniques to enable IP to improve the quality of service it can provide, while retaining the characteristics that has enabled it to become the dominant networking protocol.

The QoS in networks should provide consistent, predictable data service delivery to satisfy customers application requirements. A lot of different mechanisms enable to provide QoS, such as DiffServ, IntServ and MPLS. These mechanisms can be used for fixed and mobile communications.

This paper discusses Channelized Reserved Services (CRS), an IP-based optical architecture based on a channel abstraction, which implements a thin link layer between IP and fiber. The channelized link layer provides critical functions missing in IP--quality of service (QoS) mechanisms, bandwidth management, and redundancy. The channelized reserved services (CRS) architecture strengthens the competitive position of service providers by leveraging the optical layer to provide services on-demand. CRS is IP centric in that it is optimized to carry IP traffic. The channelized link layer includes functions not handled by IP such as QoS mechanisms, bandwidth management and protection switching. Global addressing and routing are handled by IP. CRS scales with advances in optical transport technologies due to its distributed architecture, which enables simple upgrading of the forwarding capacity of a network without necessarily replacing routers or upgrading the fiber infrastructure. Further, a spatial reuse mechanism can provide high utilization of the available transmission capacity. CRS supports high service availability in that it is engineering with restoration features such as automatic and fast protection switching.

Compared with classical applications, advanced distributed applications such as multimedia applications or distributed interactive simulation (DIS) have both new features and new constraints in terms of communication requirements. To tackle these needs, several projects have been initiated within the Internet community to design a new generation of protocols for the future Internet's architec-ture. The research reported here deals with the design of a communication architecture with guaranteed end to end quality of service (QoS) for DIS-like applications in an IPv6 environment providing 'differentiated' services. The paper is composed of three major parts: the first one presents features and requirement of DIS applications; the second part exposes the design principles of the pro-posed architecture and the networking platform on which it has been developed; the third part presents the experimental measures defined for the validation of the IP QoS management mechanisms, and then proposes and discusses a possible configuration of the architecture allowing to distribute DIS applications with respect to their communication requirements.

A lot of research efforts have recently focused on RSVP and the DiffServ architecture to provide end-to-end Quality of Service in the Internet. This paper presents a multi-layer bandwidth broker architecture, based on the DiffServ paradigm that uses RSVP as a user signaling protocol. The paper describes the necessary operations performed to support RSVP, both in terms of the architecture (functionality) and in terms of QoS management (mapping of the traffic descriptors of RSVP and the new architecture). The purpose is to step into design and implementation details that are not covered by RFCs, and to explain in brief what are the actual design decisions adopted by the authors.

Internet Service Providers need to offer Quality of Service (QoS) to fulfil the requirements of applications of their customers. Moreover, in a competitive market environment costs must be low. The selected service model must be effective and low in complexity, but it should still provide high quality and service differentiation, that the current Internet is not yet capable to support. The Differentiated Services (DiffServ) Architecture has been proposed for enabling a range of different Classes of Service (CoS). In the EURESCOM project P1006 several European service providers co-operated to examine various aspects involved in the introduction of service differentiation using the DiffServ approach. The project explored a set of service models for Expedited Forwarding (EF) and Assured Forwarding (AF) and identified requirements for network nodes. Besides, we addressed also measurement issues, charging and accounting issues. Special attention has been devoted to requirements of elastic traffic that adapts its sending rate to congestion state and available bandwidth. QoS mechanisms must prove Transmission Control Protocol (TCP) friendliness. TCP performance degrades under multiple losses. Since RED based queue management may still cause multiple discards, a modified marking scheme called Capped Leaky Bucket is proposed to improve the performance of elastic applications.

Recently a large number of approaches to service differentiation have appeared in the literature. Most of these approaches are based on the Differentiated Services (DiffServ) Architecture proposed by IETF and they have been evaluated in a number of independent studies. It is widely acknowledged that the Differentiated Services approach provides proper service differentiation in well-provisioned networks under normal traffic conditions. However DiffServ may fail to provide proper service differentiation in the presence of extreme network conditions such as congestion or in under-provisioned networks, resulting in unfair service degradation and unpredictable traffic behavior. We believe that the main reasons for the failure of service differentiation under extreme network conditions are static per-aggregate resource allocation and the lack of the specification of traffic behavior during the congestion. We propose to explicitly define traffic behavior through so called degradation policies. Degradation policies would specify how much of the network resources each traffic class could receive under different levels of congestion. In particular, the degradation policy would specify traffic behavior in terms of the performance parameters like delay, loss, or throughput. Such an approach to service differentiation allows dynamic allocation of traffic resources while maintaining fairness and predictability of the traffic behavior under all network conditions.

As Internet is rapidly growing and receiving traffic from multimedia applications that are sensitive to available bandwidth and delay experienced in the network, there is a strong need for quality of service (QoS) support. The Integrated and Differentiated Service models are two approaches for adding QoS to Internet. The Assured Service is an end-to-end service based on the Differentiated Service architecture. In this paper, we study and compare the performance of three profiler mechanisms to provide the guaranties of an Internet Assured Service. Two of them, TSW and Leaky Bucket are the most commonly used, and the third is a new Counter Based profiler, which is proposed in this paper. The study is done by simulation employing TCP RENO sources.

Multiplexing both responsive and unresponsive flows, namely TCP and UDP packets, in a single FIFO queue leads both flows into having an adverse effect on each other. In order to improve characteristics of queueing delay for UDP flows, we propose active buffer management schemes in diffserv network. We focus on MRED (Multiple RED) with HOL (head-of-line) priority control as buffer management scheme. We investigate the effectiveness of the scheme by means of simulations using ns version 2. Our simulation model has a dumbbell topology, which consists of ten sender nodes, ten receiver nodes and two routers. The five of ten senders generate CBR traffic, which will be sent by UDP, and the other five ones send bulk-data with TCP. The five CBR and the five TCP traffic share a bottleneck link between two routers. In case of FIFO, MRED with FIFO, and MRED with HOL schemes, by evaluating packet transmission delay and loss probability of CBR packet and TCP throughput, we examine how performance measures are affected by some parameters such as bandwidth of bottleneck link and thresholds of MRED. From our simulation results, MRED with FIFO is better than FIFO, but its performance is sensitive to thresholds of MRED with FIFO. On the other hand, MRED with HOL keeps stable rather than MRED in our simulation scenarios. We conclude that MRED with HOL scheme leads to low delay, low jitter, low drop probability for CBR flows as well as increases TCP throughput.

RED (Random Early Detection) is the most popular active queue management algorithm, although it has some weaknesses. Recently, another active queue management algorithm, BLUE, was proposed and shown that it is more successful in controlling the queue length when high number of flows are active on ECN (Explicit Congestion Notification) capable networks. In this paper, RED and BLUE algorithms are evaluated for different levels of RTTs, with/without ECN support. It is shown that BLUE on ECN incapable networks is not as successful as on ECN capable networks. Differentiated Services architecture suggests that RIO (Red with In and Out) style queue management algorithms are to be used on each AF (Assured Forwarding) queue to offer different levels of services for different priorities at each AF class. Inspired of BLUE's success over RED on ECN capable networks, we developed a simple alternative to RIO, BIO (BLUE with In and Out). BIO, which runs two different BLUE algorithms for in and out packets, was expected to achieve lower loss rates while maximizing link utilization for high number of active flows on AF queues. However, due to the self-configuring architecture of the algorithm, it is observed that BIO marks packets too aggressively and degrades utilization. In this paper, the properties of BIO are also explained and the results obtained are justified.

In this paper we propose a new scheme Statistical Expedited Forwarding to support a new class of service that guarantees bounded delay and bounded packet loss performance. The goal of the proposed scheme is to provide the Integrated Services type of performance guarantees, but without maintaining per-flow state in the network core. The target applications are both delay-sensitive and loss-sensitive like IP telephony, video applications and emerging on-line electronic business-related applications. In order to guarantee delay and loss bounds, the effective resource (consisting of bandwidth and buffer) is assigned to each class according to the given QoS requirement. In addition, we further study the fairness among flows in a class. The experiments on several voice and video applications show the strength of the proposed scheme.

Videoconferencing involves the transmission between the participants of different traffic flows, including such real-time flows as audio and video. In an IP internet without QoS support, the performance of real-time applications can be greatly impacted by network congestion, making them impractical. Recent advances in router technology now make it possible to protect real-time applications from the negative consequences of congestion, permitting the peaceful coexistence on the same internet of real-time and best-effort applications. We present an experimental study of the use of DiffServ technology to protect videoconferencing traffic from the congestion caused by best-effort traffic. The confirmation of the effectiveness of this protection is given by several tests, where we study the performance of point-to-point desktop videoconferencing sessions in the presence of controlled rate cross traffic. The value of the QoS parameters: delay, jitter, packet loss rate and throughput for audio and video media were obtained from the statistical data generated by the RTCP protocol. The results are immediately applicable to the use of desktop videoconferencing service in a corporate network, where this traffic can be transported together with the traditional data applications, since the traffic generated by the former will not disturb the running of the latter, and vice-versa.

There are four different levels of QoS traffic classes stipulated in UMTS specifications, which UMTS network need to support. These traffic classes are: Conversational, Streaming, Interactive, and Background. The main distinguishing factor between these classes is how delay sensitive the traffic is. Video stream class of traffic for a broadband wireless network has lot of challenges to overcome; principal among them is the preservation of time delay variation between information entities within a flow. In this paper we model video stream traffic over an UMTS network and examine end-to-end delay variation. Using the results we discuss the architectural implications of radio access network (UTRAN), core network, and the video server on the overall delay. In addition we also analyze the performance impacts of the underlying transport and network layer protocols.

The Mojette transform is a discrete projector generating bins from ixels (information elements) values. The initial bitstream is first rearranged into a 2D or n-dimensional box of ixels. Each bin value (belonging to a (n-1)-dimensional projection) is computed as the XOR addition of ixels belonging to a discrete line which fix the projection direction. The major advantage of this transform is the linear complexity (both in the number of ixels and the number of projections) for encoding and decoding. Each packet contains a projection, and additional projections can be computed without slowing the network flow. The size of this n-dimensional buffer is tuned according to the real- time constraints and to the desired packet size.

This paper discusses a technique which allows conversion of a variable-bit-rate MPEG stream at an input bit rate to a constant-bit-rate MPEG video at a specified target bit rate. This allows researchers to use real compressed VBR data to pseudo-generate a CBR stream without having the over-head of re-capturing the video sequence. Thus, using our technique researchers can use a relatively inexpensive video capture board to digitize video data and then statistically generate a variety of frame pattern and bit-rates for CBR streams. Our results will show that one can capture the long-term behavior of the video data while generating a CBR stream at a specified bit-rate. We will show experiments of our approach when applied to bandwidth smoothing and autocorrelation functions.

The third generation networks and services present opportunities to offer multimedia applications and services that meet end-to-end quality of service requirements. In this article, we present UMTS QoS architecture and its requirements. This includes the definition of QoS parameters, traffic classes, the end-to-end data delivery model, and the mapping of end-to-end services to the services provided by the network elements of the UMTS. End-to-end QoS of a user flow is achieved by the combination of the QoS control over UMTS Domain and the IP core Network. In the Third Generation Wireless network, UMTS bearer service manager is responsible to manage radio and transport resources to QoS-enabled applications. The UMTS bearer service consists of the Radio Access Bearer Service between Mobile Terminal and SGSN and Core Network bearer service between SGSN and GGSN. The Radio Access Bearer Service is further realized by the Radio Bearer Service (mostly air interface) and Iu bearer service. For the 3G air interface, one can provide differentiated QoS via intelligent burst allocation scheme, adaptive spreading factor control and weighted fair queueing scheduling algorithms. Next, we discuss the requirements for the transport technologies in the radio access network to provide differentiated QoS to multiple classes of traffic. We discuss both ATM based and IP based transport solutions. Last but not least, we discuss how QoS mechanism is provided in the core network to ensure e2e quality of service requirements. We discuss how mobile terminals that use RSVP as QoS signaling mechanisms can be are supported in the 3G network which may implement only IETF diffserv mechanism. . We discuss how one can map UMTS QoS classes with IETF diffserv code points. We also discuss 2G/3G handover scenarios and how the 2G/3G QoS parameters can be mapped.

TCP/IP protocols make minimal assumptions about the capabilities of a link layer used to carry IP packets between nodes. The recent evolution of IP-based mobile and wireless communication has driven the design of wireless link layers and new features, including support for different QoS levels, are being included in the new link layer designs. The current IP QoS architectures are based purely on IP-layer decision making, packet buffering and scheduling through a single link-layer service access point. However, mobile wireless networks pose new challenges for IP QoS architecture implementations. In this paper, we introduce a generic convergence layer interface for controlling the capabilities of wireless network interfaces. Emphasis is on discovering the link layer functionalities and mapping IP QoS mechanisms to the available link layer features. Issues such as packet buffering and scheduling are discussed.

With the increased demand on quality-of-service (QoS) guarantees, the methods used to design network elements are changing. Service models such as integrated services (IntServ), differentiated services (DiffServ), asynchronous transfer mode (ATM) service classes or multi protocol label switching (MPLS) resulted in research activities dealing with network QoS requirements. However, protocols based on many of these service models have limited implementations. Some of the reasons of these limitations are inadequate network interfaces and node services. With the development of programmable and active networks, it is possible to implement protocols and node service components that better meet the specifications of service models. This paper presents active resource control and management techniques for the realization of QoS guarantees on ATM networks. Using programmable ATM switch interfaces, we propose a set of mechanisms for the control and management of accesses to resources available at an ATM switch. The principal objective of the proposed mechanisms is to enable the implementation of flexible node and network service supports with QoS requirements, while maintaining the consistency and the integrity of resources available at nodes.

Due to the fundamental satellite system characteristics such as global coverage, broadcast nature, and bandwidth on demand, satellite systems are excellent candidates for providing high data rate Internet access and global connectivity accommodating the a wide variety of applications. Provisioning of quality of service (QoS) within the advanced satellite system is the critical requirement. Congestion remains the main obstacle to Quality of Service (QoS) on the Internet. Explicit Congestion Notification (ECN) is the only mechanism that delivers explicit congestion signals to the source. So improving the ECN feedback is essential for the future data and satellite networks and their QoS guarantees. In this paper we present a new traffic management scheme based on an enhanced ECN mechanism. In particular we used the mark-front strategy instead of the mark-tail one in ECN. The main advantage of mark-front strategy is that it sends faster feedback information about the congestion and consequently enables faster reaction from the source. Our analysis and simulation results show that the mark-front strategy of ECN provides better link-efficiency, fairness among users, lesser buffer requirements and much less losses than the mark-tail strategy.